Resource Management in Big Data Environments
A scalable network backend with real-time data processing and machine learning optimizes resource allocation and automation in cellular networks, addressing inefficiencies in traditional systems by integrating disparate data sources and enhancing network performance and customer satisfaction.
Patent Information
- Application Number
- JP2022205693
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2015-07-15
- Filing Date
- 2022-12-22
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2035-07-16
AI Technical Summary
Traditional cellular networks face challenges in efficiently managing radio resources due to legacy architectures, disparate data sources, and evolving customer demands, leading to increased operational costs and customer churn, with limited automation and integration of context information across network elements.
A scalable network backend utilizing an open architecture that combines and processes data from disparate sources in real-time, employing machine learning to optimize resource allocation and automate network operations, with a centralized measurement results megastore to share context information across network elements.
Enhances network performance by optimizing resource allocation, reducing operational costs, and improving customer satisfaction through automated problem detection and resolution, thereby minimizing subscriber churn.
Smart Images

Figure 0007588631000001 
Figure 0007588631000002 
Figure 0007588631000003
Abstract
Description
[Technical field]
[0001] (CROSS REFERENCE TO RELATED PATENT APPLICATIONS) This patent application is a continuation of U.S. Provisional Application No. 62 / 025,453, entitled “Intelligent Automatic Troubleshooting for Cellular Networks,” filed July 16, 2014; U.S. Provisional Application No. 62 / 025,958, entitled “Radio Resource Management with Big Data Feeds,” filed July 17, 2014; U.S. Provisional Application No. 62 / 025,961, entitled “Recommendation Engine for Network Insights,” filed July 17, 2014; U.S. Provisional Application No. 62 / 025,441, entitled “Automatic Healthcare Monitoring System Leveraging Cellular Data Systems,” filed July 16, 2014; and U.S. Provisional Application No. 62 / 025,513, entitled “Big Data Machine Learning Use Cases,” filed July 15, 2015, which applications are incorporated herein by reference in their entireties. This application claims priority to U.S. Provisional Application No. 62 / 193,002, entitled "Patent Cases." [Background technology]
[0002]
[0002] Currently, there is a revolution with the emergence of big data techniques. While traditional data storage technologies such as relational databases have not been performant enough for large amounts of data, the emergence of alternative cloud-based data stores, along with parallel processing algorithms such as map-reduce, have made big data practical, performant, and cost-effective. Furthermore, high performance processing, including for large amounts of data, has recently evolved that now allows real-time or near real-time analysis of data. One example includes Spark, which provides such processing for Hadoop and also leverages in-memory computation. Because big data provides statistically significant amounts of data to analyze, big data lends itself well to existing machine learning techniques.
[0003]
[0003] Big data and machine learning techniques may be applied in a wide range of areas. One example is wireless communication, which is undergoing constant and considerable transformation led by the constant development of new network technologies and services, and ushered in by the smartphone revolution and its countless applications. In this regard, wireless systems, such as cellular systems, are constantly evolving to face increasing traffic demands, as well as performance requirements imposed by these applications. The evolution of wireless communication systems generally occurs in separate blocks, such as improving performance at the radio level or improving the architecture of the core network. Although wireless communication has improved in the past decade, parts of the back-end systems supporting the network may still include legacy architectures, such as 3G and 2G networks.
[0004]
[0004] This creates tremendous operational challenges for network operators who need to grow their systems sustainably. Due to the complexities associated with the use of different types of wireless devices, different types of data services, and heterogeneous radio access networks, the cost of customer care has increased significantly. Furthermore, there has been little progress in improving various communication systems that utilize contextual information from other systems and network elements in the network. To keep up with ever-changing and demanding network performance requirements, wireless providers need to transform their maintenance cost structures while having the flexibility to cost-effectively address ever-evolving customer demands. The variability of network requirements, evolving customer needs, and the need to continue cost-effectively present challenges to traditional approaches. This disclosure is written with respect to these and other considerations. Summary of the Invention
[0005]
[0005] The drawings illustrate, by way of example only, and not by way of limitation, one or more implementations consistent with the present teachings. In the drawings, like reference numbers refer to the same or similar elements. [Brief description of the drawings]
[0006] [Figure 1]
[0006] FIG. 1 illustrates an example network architecture for implementing a radio resource management architecture. [Diagram 2]
[0007] FIG. 2 is an example high level interface block diagram of elements of a radio access network. [Diagram 3]
[0008] FIG. 2 is a block diagram illustrating a Radio Resource Management (RRM) module that interacts with a data framework consistent with an example embodiment. [Figure 4]
[0009] 1 is a flow diagram of an example process for modified RRM utilizing big data records. [Diagram 5]
[0010] 11 is a flow diagram of an example process for adjusting a connection between an eNodeB and an EU based on notifications from a measurement results megastore. [Figure 6]
[0011] FIG. 1 is a block diagram of a data framework consistent with an illustrative embodiment; [Figure 7]
[0012] FIG. 1 illustrates an example model for root cause identification. [Figure 8]
[0013] 1 is a flow diagram of an example process for maintaining radio access network performance. [Figure 9]
[0014] FIG. 13 illustrates an example user interface of a terminal coupled to a Measurement Results Megastore. [Figure 10]
[0015] FIG. 1 illustrates a high-level diagram of a system including remote health monitoring, consistent with an example embodiment. [Figure 11]
[0016] FIG. 1 is a block diagram of a data framework for a monitoring system consistent with an illustrative embodiment. [Figure 12]
[0017] 1 is a flow diagram of an example process for identifying a health condition using big data records. [Figure 13]
[0018] FIG. 1 illustrates a network or host computer. [Figure 14]
[0019] FIG. 1 illustrates a computer having user interface elements. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0007]
[0020] The present disclosure generally relates to a method and system for optimizing network elements. A scalable network backend is provided based on an open architecture that provides simplified access to data and improves automation. Operational efficiency and overall cost of network operation are reduced through automation of functions and rationalization of hardware and software resources. The use of big data techniques complemented with machine learning approaches simplifies and automates complex tasks, resulting in more efficient network operation, improved customer quality, and reduced subscriber churn.
[0008]
[0021] In one aspect, the system described herein provides an open platform where multiple data feeds from disparate sources are combined and processed in real time. In one embodiment, the system performs continuous monitoring and transmission of monitored data from each network node, which may be synthesized at a centralized remote server. The synthesized data retrieved from the disparate sources is analyzed for possible problems and relevant problem solving techniques are provided.
[0009]
[0022] In one aspect, data from disparate sources associated with a first network element is received by a processing layer. Contextual information associated with the first network element and other network elements is retrieved from a measurement results megastore. The data from the disparate sources and the contextual information are analyzed by an intelligence layer. When the intelligence layer determines that a predefined condition is met or a predefined threshold is exceeded, a notification is provided to the first network element.
[0010] (Example System Architecture)
[0023] FIG. 1 illustrates an example network architecture for implementing a radio resource management architecture. FIG. 1 illustrates a mobile communications network 100 that may be operated by a carrier or service provider to provide a wide range of mobile communications services and auxiliary services or functions to its subscriber customers and associated mobile device users. Elements generally indicated by reference numeral 100 are elements of a mobile communications network and are operated by or for a carrier. The mobile communications network 100 provides communications between user equipment (UE) 102a to 102d, as well as communications for the UEs with networks and stations outside the mobile communications network 100.
[0011]
[0024] For later discussion, several UEs are seen in the drawing, representing some examples of devices that can receive various services via the mobile communications network 122. Today, UEs commonly take the form of portable handsets, smartphones, tablet computers, personal digital assistants (PDAs), and smart watches, although they may also be implemented in other form factors, including consumer electronic devices, business electronic devices, and medical electronic devices such as wearable personal health monitor 102d. The mobile communications network 100 enables users (e.g., customers or subscribers) of the UEs to initiate communications and receive information from the Internet 120.
[0012]
[0025] The mobile communications network 100 is typically implemented by several interconnected networks. Thus, the overall network 100 may include several Radio Access Networks (RANs) 122, as well as wide area terrestrial networks that interconnect the several RANs, and wide area networks (WANs) that interconnect the wide area terrestrial networks to core network elements such as Multimedia Messaging Service Centers (MMSCs). A regional portion of the network 100, such as the regional portion serving UEs 102b and 102c, may include one or more RANs (represented by the wireless communications network 122) and regional circuits, and / or packet-switched networks and associated signaling network facilities.
[0013]
[0026] The physical elements of the RAN 122 operated by one of the mobile service providers or carriers include several base stations, represented in the example of FIG. 1 by eNodeB 119 nodes. Such eNodeB 119 nodes may include a base transceiver system (BTS) that communicates with one or more UEs (102a through 102d) via an antenna system at the node's site and over an airlink when the UEs are within range. Each eNodeB 119 node may include a BTS coupled to several antennas mounted on a radio tower within a coverage area, often referred to as a "cell." The BTS is the part of the wireless network through which the eNodeB node 119 transmits RF signals to and receives RF signals from the UEs it currently serves. The radio access network 122 carries user communications for UEs between each eNodeB 119 node and other elements with which or through which the UE communicates. In various embodiments, the radios transmitting communication signals (to one or more UEs) at each eNodeB may adjust power output, bandwidth, etc. based on context information received from the Measurement Results Megastore 140, and / or alerts received from the Monitoring Server 130, discussed in more detail below. Individual elements, such as switches and / or routers, that form the Radio Access Network 122 are omitted here for simplicity but will be discussed where relevant in connection with FIG.
[0014]
[0027] The radio access network portion 122 of the mobile communications network 100 connects to a public packet data communications network, such as the network commonly referred to as the "Internet," shown at 120. Packet communications via the radio access network 122 and the Internet 120 can support various services throughout the network 100, such as providing contextual information from a measurement megastore and alerts from a monitoring server 130 to various network elements of the radio access network 122, as well as providing text and multimedia messaging, e-mail, web surfing or browsing, programming, media downloading (including audio and video), and the like, to the UEs. For example, the UEs 102a through 102d could receive messages from and send messages to user terminal devices, such as personal computers, either directly (peer-to-peer) or via various servers (not separately shown).
[0015]
[0028] For the purposes of this discussion, it should be noted that many of the network messages and alerts discussed herein are sent to and received from various UEs using the Radio Access Network 122 and elements of the Radio Access Network 122. The Evolved Packet Core (EPC) of the Radio Access Network 122 uses the concept of Evolved Packet System (EPS) bearer channels to route IP traffic from the Gateway 107 in the Radio Access Network 122 to the UEs. A bearer channel is an IP packet flow with a defined Quality of Service (QoS) between the Packet Data Network (PDN) Gateway (PGW) 107 and the UEs (e.g., 13a through 13e).
[0016]
[0029] The eNodeBs 119 are interconnected with each other by an interface commonly known as "X2". Communication between the eNodeBs includes Radio Resource Management (RRM), which covers all radio bearer related functions such as radio bearer control, radio admission control, radio mobility control, scheduling, dynamic allocation of resources to UEs in both uplink and downlink, etc. In one aspect, the RRM can communicate with the Measurement Results Megastore 140 and / or receive messages from the Monitoring Server 130 to optimize its control functions.
[0017]
[0030] In one embodiment, the network 100 includes a measurement megastore server 140 coupled for communication over the Internet 120. The measurement megastore server stores context information (which may be from various sources) relevant to creating a more efficient connection (e.g., power consumption and / or speed) between the UE and the radio access network 122, as well as optimizing the components of the radio access network 122. For simplicity, the UE and the components of the radio access network are collectively referred to herein as network elements. In various embodiments, the context information provided by the measurement megastore server 140 can include different types of information. For example, the context information can include application level information extracted via Deep Packet Inspection (DPI) received from one or more radio access networks 122. The context information at the server 140 can include information of the UE terminal category (e.g., smartphone, tablet, laptop, personal health monitoring device, etc.). The context information may also relate to the capabilities of the network element, such as maximum transmission bit rate and whether it is configured to be a low power device (configured for low power as compared to high performance). In one embodiment, the context information may provide a recently stored power status of the network element, which may be used to optimize the connection between the radio access network 122 and the respective UE.
[0018]
[0031] In response, the measurement megastore server 140 stores context information including disparate data from various sources. The data stored in the server 140 includes information that can be used by the network elements of the wireless communication network 122 to provide coordinated optimization of the resources of each network element, respectively. Alternatively, or in addition, the monitoring server 130 can retrieve context information including information from the entire network 100 from the measurement megastore 140 and analyze this context information together with data from disparate sources related to the network elements to determine whether one or more predefined conditions are met that trigger a message to the respective network element to optimize its performance in the context of the entire network 100. In contrast, prior art cellular communication networks can simply utilize metrics obtained locally from a small subset of network elements, such as network elements in a core network from the same vendor or network elements deployed to implement specific functions in the network, to determine how to optimally allocate resources for each communication link of the UE 102 to the radio access network 122.
[0019]
[0032] For example, to establish a connection between a UE and an eNodeB, a prior art radio scheduler (at the eNodeB) can only consider locally accessible information, such as the amount of data in each UE's buffer, the identified radio quality, device capabilities, etc., to determine the number of turns the scheduler should allocate to each link between the UE and the corresponding eNodeB. Even in more advanced prior art systems that can use deep packet inspection (DPI) to assign a different quality level to a Skype video call compared to, for example, a File Transfer Protocol (FTP) transfer, data is extracted from a specific subsystem (i.e., the wireless communications network) and is not shared between different network elements. However, this extracted data is limited and the resources allocated to a connection between an eNodeB and a corresponding UE are at best guesswork. For example, in prior art systems, even if the radio scheduler of the eNodeB 119 is able to determine the terminal category of the terminal, the scheduler may still not know that the UE (e.g., 102d) is a machine to machine (M2M) module that needs to conserve battery life as much as possible.
[0020]
[0033] In contrast, in one aspect, the system described herein enables sharing of information across multiple eNodeBs 119 by storing context information at a Measurement Megastore Server 140, which improves the capacity of the wireless communication network 122 by, for example, allocating types of resources to connections between UEs (e.g., 102d) and their corresponding eNodeBs 119. Thus, wireless communication between the UE 102 and the radio access network 122 is better coordinated by utilizing context information such as customer type, data plan, application type, device type, etc., provided by the Measurement Megastore 140 via the Internet 140.
[0021]
[0034] The monitoring server 130 and the measurement megastore 140 can be implemented in a distributed manner on several similar platforms to distribute the processing load and / or provide additional functionality. For example, in one embodiment, there may be a remote health (e.g., for humans) monitoring server 150, which is similar in configuration to the measurement megastore 140, but may be dedicated to a form of remote health monitoring. In this regard, the system 100 provides an open platform where various UEs, such as the personal health monitor 102d, and multiple data feeds from heterogeneous systems, such as other networks and servers, are combined and processed for millions of individuals in real time. Thus, the functionality described herein for each of the servers 130, 140, and 150 may also be provided by one or more different servers. The actual operation of the system 100 is described later by way of example.
[0022]
[0035] 2 illustrates an example high-level interface block diagram of logical nodes of the radio access network 122. For example, the System Architecture Evolution (SAE) 200 is a core network architecture of the 3rd Generation Partnership Project (3GPP) Long Term Evolution (LTE) wireless communications standard. The SAE 200 has a flat all-IP architecture with separation of control plane traffic and user plane traffic. The SAE 200 system includes an eNodeB 219, a Mobility Management Entity (MME) 202, a Home Subscriber Server (HSS) 204, and a Policy Control and Charging Rules Function (PCRF) 206, which are part of the mobile traffic network 122 and each of which is a network element. The eNodeB 219 can be viewed as an enhanced base station that serves as an interface between the UE 202 and the radio access network 122. The MME 202 is the control node for the communication access network. For example, the MME 202 is used for idle mode mobile device tracking and paging procedures, including retransmission. The MME 202 also provides bearer channel activation / deactivation processes and the selection of a serving gateway (SGW) 212 for the UE 202e.
[0023]
[0036] The Home Subscriber Server (HSS) 204 is a central database containing user-related and subscription-related information. The HSS 204 provides functions such as mobility management, call and session establishment support, user authentication, access authorization, etc. For example, the HSS stores the Mobile Station International Subscriber Directory Number (MSISDN), SIM card identification information, etc.
[0024]
[0037] A Packet Data Network (PDN) Gateway (PGW) 107 provides connectivity from the UE 202 to external packet data networks. A Policy Control and Charging Rules Function (PCRF) 206 provides policy control decision making and control of flow-based charging functions in the Policy Control Enforcement Function (PCEF), which resides in the PGW 107. The PCRF 206 determines how certain data flows are handled in the PCEF and also provides QoS authorization to ensure that this is in accordance with the user's subscription profile.
[0025]
[0038] The Serving Gateway (SGW) 212 routes and forwards data packets to the mobile 202 through the MME 202. For example, the data packets may include notification messages that an account threshold is exceeded. The SGW 212 manages and stores mobile device parameters for IP bearer channel services, including network internal routing information. Each of these and other network elements may be adjusted in performance based on context data from the Measurement Results Megastore 140 and monitoring server 130 alerts / messages.
[0026] (Radio Resource Management (RRM))
[0039] 3 is a block diagram illustrating a radio resource management (RRM) module that interacts with a data framework consistent with an example embodiment. The eNodeB 319 serves as a base station between one or more UEs (e.g., UE 302) and the wireless communication network. In the example system 300, each eNodeB 330 includes a scheduler 322 configured to provide priority (e.g., allocate number of turns) for data transmission between the UE 302 and the wireless communication network. The eNodeB 319 can include a link adaptation unit 324 configured to adjust the transmission mode to the channel conditions, for example, by adjusting modulation, encoding, and selecting an appropriate time for transmission.
[0027]
[0040] The eNodeB 319 may include a flow control unit 326 configured to control a flow of operation of the eNodeB 319. The flow control unit 326 may provide admission control, coordination of security signaling, compression, etc. The eNodeB includes an antenna 330 configured to communicate with various UEs (e.g., UE 302) within its wireless communication range.
[0028]
[0041] The various components of the eNodeB 319 together include an RRM module 328 that is configured to communicate with a data framework 370 via an open Application Program Interface (API) 334. As shown in the system 300, the data framework can include a measurement megastore 340, an intelligence layer 350, and a data processing layer 352. For example, the API 334 can enable communication between the eNodeB 319 via the wireless communication network 122 to the measurement megastore 340 via the Internet 120 (of FIG. 2). The data processing layer 352 can also receive localized information from the RRM 328.
[0029]
[0042] As discussed above, the measurement results megastore 340 is configured to store context information that may be shared by the eNodeB 319 (and other eNodeBs configured to receive information from the measurement results megastore 140) from various network elements, such as RRM modules of different eNodeBs, to better coordinate respective network elements of wireless communications over the wireless communications network 122 of FIG. 1. Such optimization may include at least one of an admission control procedure, a scheduler procedure, a link adaptation procedure, and a power control procedure of the respective eNodeB.
[0030]
[0043] In an embodiment, data is collected by a data processing layer 352 of the system 300 to populate the context information in the measurement megastore 340. The data collected by the processing layer can be from a variety of sources, including data from elements of the wireless communication network 122, such as an RRM module coupled to the wireless communication network 122, as well as other servers (e.g., 130, 150) of the system 100. The types of information received by the data processing layer 352 can include subscriber radio trace (e.g., location) information provided by the UE, operations subsystem (OSS) information, customer care record information, billing information, application performance monitoring report information, and other information, including relevant information from the Internet. In one embodiment, at least a portion of this information is provided by the measurement megastore 340.
[0031]
[0044] In one embodiment, the data includes information from deep packet inspection (DPI) provided by various eNodeBs. The data processing layer 352 may be performed in real-time and / or in batches. The data may be provided to the network elements (e.g., the RRM module 328 in the eNodeB 319) via queries performed by the network elements in real-time, or may be pushed to the RRM module at predefined time intervals or in response to notifications generated by the intelligence layer 350.
[0032]
[0045] In one embodiment, the intelligence layer 350 may be configured to receive information from the data processing layer 352 and determine whether one or more predefined conditions are met. These conditions may be provided by a rule set stored in the memory of the intelligence layer. When the predefined conditions are met, the intelligence layer may generate an associated alert. For example, the intelligence layer 350 may analyze patterns in the data provided by the data processing layer 352 for any anomalies that may indicate a problem in the RRM. In this regard, the intelligence layer 350 may trigger notifications to other RRM modules to prevent and / or mitigate the problem so that it does not spread. An alert may simply be a notification (e.g., a proposal to act) or an active control signal to change the RRM methodology of one or more modules that are part of the system 300.
[0033]
[0046] The intelligence layer 350 typically includes a probability tree that stores the probability that a particular cause and / or event was the root cause of a network event, such as a dropped call. An implementation of a probability tree is described in further detail with reference to Figure 7. Note that a node with an intelligence layer 350 need not have the same probability tree as other nodes. In this manner, nodes may also be partitioned according to scenarios and / or use cases.
[0034]
[0047] For example, the data processing layer 352 may receive information from the first RRM module indicating a problem. The data received from the first RRM module may, for example, indicate that the SGW is overloaded using a signaling message. Upon identifying a problem with the first SGW, the intelligence layer 350 may retrieve context information related to the network element and other network elements from the measurement results megastore. Upon determining that a condition is met or a predefined threshold is exceeded, the intelligence layer 350 may issue a notification to other RRM modules configured to receive messages from the data framework 370. In one embodiment, the first RRM module also receives alerts from the data framework 370 generated by the intelligence layer 350. In this regard, the other eNodeB modules may buffer or drop communication requests heading to their systems, respectively, to prevent the problem that affected the first RRM module.
[0035]
[0048] Thus, the context information in the Measurement Results Megastore 340 may be made available to the RRM module by the Measurement Results Megastore 340 in response to a request (e.g., a query) from the RRM module, or may be provided as a notification from the Data Framework 370.
[0036]
[0049] With the above overview of the system, it may now be helpful to consider a high-level discussion of an example operation of the system 300. To that end, FIG. 4 illustrates a high-level flow chart of a data framework 430 and an RRM module 402 and their interaction with each other. The RRM module 402 may be one of several RRM modules that are configured to receive data from the data framework 430. The data framework 430 includes context information collected from multiple sources, which is used by the RRM module 402 to adjust resources of the RRM module (e.g., optimize the connection between the eNodeB and the UE).
[0037]
[0050] Starting with the data framework 430, in step 434, relevant data is collected from various sources into a common data framework. This data is collected by the data processing layer in real time and / or in batches. The collected data can be from various sources, which can include different components of the wireless communication network 122, including separate RRM modules configured to operate on the wireless communication network 122. For example, each RRM module can provide radio resources based on its local knowledge of the wireless communication network 122 of the system 100 of FIG. 1 and its corresponding UE. Sources of the collected data can also include other servers of the system 100 and the Internet. The types of information received by the data framework 430 (via the processing layer) can include subscriber radio trace (e.g., location) information provided by the UE, operation subsystem (OSS) information, customer care record information, billing information, application performance monitoring report information, and other information, including relevant information from the Internet. In one embodiment, the data retrieved by the data processing layer ultimately becomes context data stored in the Measurement Results Megastore Server 340, which may be used and retrieved by other network elements or data processing layers.
[0038]
[0051] In one embodiment, data collected by the data framework 430 from various sources is provided to the RRM module 402 in step 436 to adjust the resources allocated to the connection between the eNodeB and the corresponding UE. The information may include data not processed by the intelligence layer, such as general information about subscribers and network nodes.
[0039]
[0052] In one embodiment, in step 438, the data collected by the data framework is further processed by an intelligence layer, which operates to analyze the data and extract patterns and / or trends therefrom. For example, in step 438, the intelligence layer may determine whether one or more thresholds are exceeded or one or more criteria are met. For example, the RRM module is off or the number of dropped calls exceeds 0.1%. In various embodiments, considerations by the intelligence layer may include user / subscriber predefined settings, subscriber limitations (e.g., exceeding a threshold data usage), plan type (e.g., higher grade plans may be given priority), UE type (e.g., devices intended to function under low power conditions are provided with connections that require less power), etc. In this regard, in step 422, the data framework 430 can provide one or more messages in the form of a special alert to the appropriate RRM module, which can include instructions on how to adjust local resources for the connection between the eNodeB and its corresponding UE, individually. For example, the appropriate RRM module to which the alert is sent from the data framework 430 can be based on sector and / or time (e.g., RRM coverage of a sports stadium at game time).
[0040]
[0053] On the RRM module 402 side, in step 406, the RRM module 402 calculates radio resources based on local knowledge, which may include the type of device (UE) with which it has established a connection, the number of UEs with which the local eNodeB is connected, traffic, power consumption, modulation quality, temperature of the eNodeB, etc. In one embodiment (not shown), this information is provided to the data framework 430 via the data processing layer 352.
[0041]
[0054] In step 406, the RRM module 402 adjusts resource allocation between the eNodeB and each UE individually based on the general information data received from the data framework 430 in step 420. As discussed above, in one embodiment, the RRM module 402 may receive one or more messages in the form of one or more special alerts (i.e., notifications) from the data framework 430. In step 408, the RRM module 402 allocates eNodeB resources to subscribers and network nodes based on the special alerts received from the data framework. Although steps 406 and 408 are shown as examples to be separate steps, in one embodiment, the resource adjustment and resource allocation may be performed by the same step (e.g., 408).
[0042]
[0055] For example, based on information received from the data framework 430 with the measurement megastore 140, the eNodeBs of the RRM module 402 can adjust connections based on power (e.g., consumption limits) requirements for each UE individually. In various examples, based on context information received from the measurement megastore 140, the eNodeBs can gradually reduce the number of UEs allowed to connect to each eNodeB to balance resources of the radio access network 122 and / or prevent failures. For example, the scheduler 322 can stop sending new session requests to the radio access network 122 upon receiving a warning from the measurement megastore 340.
[0043]
[0056] Other corrective actions may include reducing the bandwidth or preventing communication with a particular UE that is exceeding a predetermined threshold or does not meet one or more criteria. In various embodiments, the constraint may be increased after a threshold time or upon receiving a message from the measurement results megastore 140 that the UE is within a predetermined threshold. In one example, the constraint may be increased when a criteria is met (e.g., a sporting event at an arena serviced by the eNodeB 319 is completed), which reduces the risk of overloading the eNodeB of the RRM module 402.
[0044]
[0057] 5 is a flow diagram of an example process for adjusting a connection between an eNodeB and an EU based on notifications from the measurement results megastore 140. The flow diagram of FIG. 5 is a specific example that considers the type of network element and stored criteria for such type of element. In step 506, the RRM module determines the radio signal quality between the UE and the eNodeB based on local knowledge. In one embodiment, in step 508, the RRM module determines an appropriate encoding format based on the determined radio signal quality based on local knowledge.
[0045]
[0058] In step 510, the RRM module communicates with and receives context information from the Measurement Megastore 140. This step may be performed as a query from the RRM module to the Measurement Megastore 140, or as a push notification from the Measurement Megastore 140 to the RRM module based on predefined criteria stored in the Measurement Megastore 140.
[0046]
[0059] In the example of FIG. 5, the RRM module can support various network operations to perform queries (e.g., over the Internet 120) to the Measurement Results Megastore 140. In this example, the network operation is to initiate and establish communication for the UE. In various embodiments, the RRM can transmit a UE device identifier and / or other indicators of wireless signal quality based on local knowledge to the Measurement Results Megastore 140. In response, in step 520, the Measurement Results Megastore 140 can indicate that the type of UE with which the RRM module is in the process of establishing a connection is intended to function at low power and that the connection between the eNodeB and the UE should be optimized for low power operation. For example, the connection may be serving an M2M customer that can benefit from a connection that uses a minimal amount of power. This scenario may be important for medical devices that are embedded personal health monitoring devices or wearable personal health monitoring devices. Thus, in step 520, the context information provided by the measurement results megastore 140 can be in the form of a notification (e.g., a warning) or instruction to the RRM module to adjust the connection so that the UE consumes a more optimized (e.g., minimal) amount of power.
[0047]
[0060] In another scenario, the indicator of radio quality determined based on local knowledge and provided to the Measurement Megastore 140 may include a battery status (e.g., low). In this regard, the Measurement Megastore 140 in cooperation with the Intelligence Layer may identify from a previously stored user configuration that the UE, which is now low on power, should be operated in a low power mode to extend its duration as much as possible. In this regard, the Intelligence Layer may issue a notification to the RRM module (at step 520) that the connection between the eNodeB and the UE should be of low power. Thus, without additional input from the Measurement Megastore 140, the RRM module would not be able to determine that the user configuration includes low power operation when a predefined condition is met (i.e., running at low power).
[0048]
[0061] In response, in step 512, the RRM modifies the coding scheme to provide a connection with the UE that provides the greatest power savings, which may include lower bandwidth operation, thereby freeing up resources in the radio access network 122 for other purposes. For example, an alert received from the intelligence layer may trigger an "event handler" in the eNodeB to perform some power saving function.
[0049] (Intelligent and Automatic Troubleshooting for Radio Access Networks)
[0062] As discussed above, in one aspect, the use of the measurements megastore 140 enables automatic detection of network problems at the subscriber level. In some scenarios, problems may be identified (and sometimes prevented) even before the customer becomes aware of the problem, which greatly improves customer relations. This advantage can be better understood in light of existing approaches to address network problems and associated customer dissatisfaction. For example, when a subscriber (e.g., a customer) experiences a network problem, a lengthy, tedious, and sometimes frustrating process is typically involved that is typically very labor intensive and requires frequent intervention of a skilled engineer to identify and resolve the issue.
[0050]
[0063] For example, a customer may first have to call and describe the symptoms to a customer care representative, who will classify the problem based on predefined criteria. An internal record may be generated and analyzed by an appropriately skilled engineer. The engineer may generate a call reliability report that includes traffic volume, number of dropped calls, reasons for dropped calls, and an overall calculated dropped call rate for the node in question. In some cases, there is an intermediate engineering triage group that can address some of the complaints, while in other cases, the case is sent to a field engineer to investigate the root of the problem. Due to the complexity of the wireless access network, a field engineer may investigate multiple systems to determine what may be the cause of the problem based on the limited analysis that the field engineer can perform from his local knowledge of the network.
[0051]
[0064] In today's cellular networks, the engineers' challenges are further exacerbated by the growing number of data sources with a significant amount of performance data collected from each network element (e.g., eNodeB). This performance data is often aggregated within time intervals such as minutes, hours, or days. Together, the data provide key performance indicators (KPIs) that are reviewed by engineers to better understand the overall health of the radio access network, detect problematic conditions, and determine when to upgrade portions of the network. Furthermore, in existing systems, it is difficult to aggregate data from disparate sources, especially when they are of different nature. For example, a report containing a trend in drop call rates in one area cannot easily be combined with a trend in churn rates. Such dedicated reports require significant processing and therefore may not be immediately available to the engineers. Even after the data is made available, the engineers still need to manually analyze large amounts of data. This manual analysis can be inefficient as well as incomplete, as the engineers may not be able to effectively understand trends from multiple data sources. The result is typically suboptimal radio access network performance.
[0052]
[0065] In contrast, in the system described herein, most (if not all) of the analysis of the network's condition is performed automatically by the data framework including the measurements megastore 140. In fact, the root cause of a problem may be determined (if not prevented) even before the customer knows about the problem. Such automation is made possible by the measurements megastore 140, which, along with the data processing and intelligence layers, is configured to detect problems at the subscriber level by relying not only on local knowledge of the network elements, but also by using data related to the entire network, and by automatically remedying the problem and / or providing notification (e.g., reporting it) to appropriate representatives and / or field engineers.
[0053]
[0066] In one embodiment, to accommodate the real-time and large volume of information being generated by a large number (e.g., millions) of subscribers, the Measurement Results Megastore 140 uses massively parallel processing technologies such as Hadoop, Storm, or Spark. In addition, a machine learning component is used to identify patterns and trends based on a predefined training set. In one embodiment, previously unknown problems are detected by a different clustering model. For example, data is grouped such that objects in the same group (i.e., cluster) are more similar to each other than objects in other groups (i.e., clusters).
[0054]
[0067] Figure 6 is a block diagram of a data framework consistent with an example embodiment. The data framework of Figure 6 is similar to the data framework of Figure 3 discussed above, but is shown again for clarity to facilitate discussion of automated troubleshooting of a radio access network. The data framework of Figure 6 includes a data processing layer 682, an intelligence layer 684 that may include programmable pattern detection rules 688, an automation layer 686 configured to store a set of programmable actions based on a problem type 690, and a measurement results megastore 680.
[0055]
[0068] The type of information received by the Data Processing Layer 682 may include, among other elements: data from customer care trouble ticket databases; network element counters and KPIs; network element (UE) locations; alert and fault information; subscriber information including subscriber plan, usage (e.g., reports / data), billing addresses, etc; handset information including handset type, OS, applications, battery status, etc; subscriber detail information records including geolocation traces, radio traces, core network traces, and IP level traces; Wi-Fi records and metrics, etc. The Data Processing Layer 682 may be performed in real-time and / or in batches. Monitoring may be performed continuously, around the clock, at predefined intervals, or upon triggers from the monitoring system intelligence layer 384. Thus, the Data Processing Layer 682 monitors detailed subscriber traces and also records fault conditions such as dropped calls, access failures, problems with Internet connectivity, etc. This data is stored in the Measurement Results Megastore 680 along with data from other network elements as contextual information that may be used by various network elements.
[0056]
[0069] The intelligence layer 684 is configured to receive information from the data processing layer 682 and pattern detection rules 688 that may be pre-programmed in the intelligence layer 684. In one embodiment, the intelligence layer 684 may include different intelligence models for different types of failures. The intelligence layer 684 is configured to identify likely causes of the problem as well as an estimate of the accuracy (error) of the error detection. In one embodiment, the accuracy of the detection is based on the frequency of occurrence of the error.
[0057]
[0070] For example, the accuracy estimate can be higher when customers may be experiencing degraded performance (disconnections, access failures, etc.) in areas identified by the intelligence layer as undergoing maintenance or known coverage deficiencies. The intelligence layer can further improve accuracy by considering customer traces, network health indicators, UE reports, etc. From disparate sources of data, the intelligence layer 684 can identify whether a disconnection is due to (1) coverage, (2) interference, (3) hardware failure, (4) software failure, etc.
[0058]
[0071] The automation layer 686 can trigger an automated response to an identified network error without relying on prior human processing of the raw data retrieved by the data processing layer 682. In one embodiment, the appropriate action for an identified problem is pre-programmed by a system administrator at the intelligence layer or at the network element node. The action can be to reset the node or send a note to a field engineer with the identified problem and a possible problem-solving technique. In other cases, the automated action can be to send a message to the subscriber informing them of the cause of the problem (e.g., there is maintenance work in progress in the area) or to provide an automatic credit to compensate for the deteriorated performance and to inform the subscriber of the same.
[0059]
[0072] In one example, if multiple subscribers are affected by a substantially similar problem in an area, the automation layer 686 can automatically send a message to all subscribers in the area to proactively remedy the problem before the subscribers are affected by the problem. In other words, when a group of other network elements in the area of a first network element determines that a predefined condition is met or exceeds a predefined threshold, a notification is automatically sent to all network elements in the group. In one embodiment, network errors identified with an accuracy below a first threshold are treated differently than network errors identified with an accuracy above a second threshold. In one embodiment, the first and second thresholds have the same value. For example, if an error is detected that is deemed to be of low accuracy, an automatic action may not be taken until additional data is received that improves the accuracy of the problem and / or the subscriber calls. However, if an error is identified with high accuracy, an automatic immediate remedial measure is undertaken.
[0060]
[0073] For example, the automation layer 686 in cooperation with the intelligence layer 684 can suggest mitigating actions when it determines that a particular subscriber is not receiving the expected quality of service. In the case of coverage issues, the intelligence layer can determine whether there is an option to use Wi-Fi calling. In this regard, the automation layer 686 can send notifications to the subscriber and care representatives to enable activation of Wi-Fi calling.
[0061]
[0074] FIG. 7 illustrates an example model for automatic root cause identification. Specifically, FIG. 7 illustrates a probability tree that stores information about likely root causes for dropped calls. Based on a large corpus of data, machine learning algorithms can determine the probability that an event, such as a dropped call, is correlated to a particular cause. Similarly, a particular cause can have its own different possible causes, each of which has its own probability. Finally, the probability of the root cause can be ascertained.
[0062]
[0075] In Figure 7, we start at the top of the probability tree with a dropped call event 702. Every event is associated with a probability P(event). For example, the probability that the detected network event is a dropped call is shown as P(yes) = 0.15 (or 15%). The probability that the detected network event is some other event is P(no) = 0.85.
[0063]
[0076] The probability tree can then be traversed to determine the most likely root cause. To do so, the probability tree is traversed to nodes that represent possible direct causes for the dropped call event 702. These may include wireless interference 704, poor coverage 706, hardware failure 708, software failure 710, or other failures 712. As with the dropped call event, each cause is associated with a probability. Because these causes are possible causes, their probabilities are expressed as conditional probabilities P(condition|event). For example, the probability that the dropped call was caused by a hardware failure 708 is shown as P(HW|DC)=0.2. Similarly, the probability that the dropped call was caused by a software failure 710 is shown as P(SW|DC)=0.3.
[0064]
[0077] Causal nodes may in turn have their own respective causes. For example, wireless interference 704 and insufficient coverage 706 may have the same cause 714. A hardware failure 708 may have different possible causes, shown here as a physical error 716 and a misconfiguration of the user equipment 718. Again, each node is associated with a respective probability, and as a possible cause, each probability is expressed as a conditional probability. Here, the probability that a physical error 716 in the hardware caused the hardware failure is shown as P(PH|HW)=0.1. The probability that misconfigured hardware caused the hardware failure 718 is shown as P(MCG|HW)=0.6.
[0065]
[0078] Note that software failure 710 can also be caused by incorrect configuration 718. Even if the probability tree traverses to a node that is also traversable from hardware failure 708, the incorrect configuration node 718 has its own probability as the cause of software failure 710, as shown as P(MCG|SW)=0.3. Software failure 710 can also be caused by having the wrong version of software 720 installed, with a conditional probability of the cause of P(WVER|SW)=0.5.
[0066]
[0079] Similarly, the other causes 712 may have their own respective causes 722 .
[0067]
[0080] A probability tree is generally populated via receiving a statistically significant corpus of data. The data is then applied to the probability tree to calculate the probabilities in the probability tree.
[0068]
[0081] When performing diagnosis, the probability tree starts with cause and / or KPI leaf nodes 724, 726, 728, 730, 732. Depending on the observed network data, the probabilities of the cause and / or KPI leaf nodes 724, 726, 728, 730, 732 are updated. For example, if an incorrect configuration is actually observed, the probability of incorrect configuration 728 is changed to 100%. If cause 724 is not observed, its probability is changed to 0%. Probabilities do not have to be only binary, i.e., 100% or 0%. Intermediate (non-0%, non-100%) probabilities may also be calculated based on a sampling or moving average of the data.
[0069]
[0082] After the leaf nodes 724, 726, 728, 730, 732 of the causes and / or KPIs are updated, the probabilities of the parent nodes are updated according to the Bayesian algorithm. Thus, the conditional probabilities of the parent nodes 714, 716, 718, 720 and 722 are updated using the child nodes as the decision as to whether the condition is met. The probabilities of the parent nodes 714, 716, 718, 720 and 722 are then updated. This process is repeated for the levels of parent nodes above until the root node 702 is reached. At this point, all probabilities in the probability tree have been updated taking into account the observed causes and / or KPIs 724, 726, 728, 730, 732.
[0070]
[0083] The possible root causes of the root node, the dropped call 702, will be revealed to the user by selecting all child nodes whose updated probabilities exceed a predefined threshold.
[0071]
[0084] After completing the analysis, the probability tree is reverted to its original state and original probabilities.
[0072]
[0085] However, over time, historical data is collected, and upon agreement by an administrator, at least a subset of the historical data can be applied to the probability tree to permanently update the probabilities in the probability tree. Subsequent analyses will then use the updated probability tree. In this manner, the probability tree learns from the historical data.
[0073]
[0086] It should be emphasized that this particular approach in machine learning is just one option. Other well-known options include entropy-based trees and naive Bayesian trees. One advantage specific to Bayesian-based trees is that the domain of causes and / or KPIs can be partitioned into subdomains without forcing a recalculation of probabilities. Accordingly, probability trees may be partitioned for different subsets of the network as desired by the administrator.
[0074]
[0087] FIG. 8 is a flow diagram of an example process for assisting engineers in maintaining radio access network performance. In step 802, data to be stored in the Measurement Megastore is automatically collected from various sources into a common data framework. In step 804, pattern recognition is performed on the data by one or more clustering algorithms. In one embodiment, in step 806, key metrics to be analyzed are preprogrammed in the Measurement Megastore or are interactively provided by an operator to the Measurement Megastore via a user interface 900 (see FIG. 9). The user interface 900 can be via a software application on a terminal or handheld device that is communicatively coupled directly or wirelessly to the Measurement Megastore 140 of FIG. 1. In various embodiments, an engineer can select (e.g., 904) the type of analysis to be performed and the level of detail to be reported, both at the network element (e.g., geographic / node level) and at the KPI level.
[0075]
[0088] There are an unlimited number of patterns that the Measurement Results Megastore can detect. For example, an area with a high churn rate may indicate that most of the subscribers are concentrated in a small geographic area. A sector experiencing degraded performance may indicate a large number of resets in a particular period of time. A large number of customers calling customer care may indicate the number of calls being routed directly to voicemail exceeding a predefined threshold. An area showing a dropped call rate that is within a predefined threshold may indicate several instances where customers have had to redial within a predefined time period, exceeding the threshold (which may indicate poor call quality). A sector may indicate congestion despite carrying traffic below the threshold. A threshold number of customers may experience a reduction in their coverage following a network upgrade. Of course, countless other scenarios are supported.
[0076]
[0089] As they are learned, in step 808, the relevant patterns continue to reside in memory, for example, in the measurements megastore 140. In one embodiment, the measurements megastore 140 displays the analysis results, including the most relevant patterns, to the engineer on the user interface 900. In another embodiment, patterns that are associated with automated solutions can trigger the execution of those automated solutions.
[0077]
[0090] In one embodiment, at step 810, an engineer may rate the results of the measurements megastore analysis. In this regard, diagram 900 illustrates, by way of example, an interface that may be used by an engineer to rate the suitability of the measurements megastore analysis (e.g., findings) by selection of stars 906 from 1 to 5. In various embodiments, the selection may be performed after the engineer has reviewed and fixed the network issue or immediately after initial display on the user interface 900.
[0078]
[0091] In one embodiment, in step 810, an engineer can request that automatic KPIs and / or notifications be issued based on the analysis. The KPIs can be stored in a measurement megastore and used as criteria. The KPIs and notifications can be used to predict events such as when an area is likely to have increased churn levels.
[0079]
[0092] In one embodiment, in step 812, the system described above may be trained, where the system automatically repeats the previous analysis based on previous similar situations and continuously generates similar KPIs and / or alerts at predetermined intervals or when a specific error is identified. In one example, the intelligence layer of the system described herein can be combined with a self-organizing network (SON) to perform automatic planning, configuration, management, optimization, and recovery of the radio access network.
[0080] (Example use cases in the medical field)
[0093] As discussed above, in one aspect, the system described herein provides an open platform where multiple data feeds from heterogeneous sources may be combined and processed in real time. Much of the prior discussion has been directed to applications in network management. However, the techniques described herein provide learning and feedback loops that may be applied to any field where there is a stream of event data, where the event data may be classified into categories that may be correlated with causes. Where probability trees may be constructed, prior techniques may be applied.
[0081]
[0094] One exemplary alternative field where this capability is particularly useful is in the medical field, which is increasingly using health monitors, such as UE 102d, that can be worn by both healthy and sick individuals to monitor the wearer's medical health status. Today, one or more such health monitors are worn and can provide a range of medical analytics about the user (sometimes referred to herein as the patient or subscriber) and can be provided to an appropriate caregiver to extract relevant information from the one or more health monitors.
[0082]
[0095] These health monitors can come in various forms, such as implanted, worn on the wrist, worn around the neck as a necklace, strapped around the chest, etc. These monitors can monitor various health functions, such as heart rate, heart rhythm, blood pressure, movement, oxygen saturation, body temperature, blood glucose levels, etc. Traditionally, each health monitor operates in isolation. For example, a health monitor that measures heart rate is mutually exclusive from a health monitor that measures blood glucose levels or movement. The acquired data points are then either provided directly to the patient on a user display, or stored in the health monitor's memory, respectively, to be read later by one or more caregivers. In some cases, the acquired data is transmitted periodically or immediately after a trigger from the user to heterogeneous receivers at one or more caregivers. Each health monitor is generally applied in response to a health condition suspected by the caregiver. Accordingly, in conventional systems, constant monitoring that provides a synthesis of information from the health monitors that identifies unsuspected health conditions is not available.
[0083]
[0096] In contrast, the system described herein may perform constant monitoring and / or constant transmission of monitored data from each health monitor, which may be combined at a centralized remote health monitoring system 150. The combined information retrieved from the heterogeneous sources is configured to analyze the data for possible diseases and provide relevant notifications.
[0084]
[0097] Reference is now made to Figure 10, which illustrates a high-level diagram of a system including remote health monitoring consistent with one example embodiment. In system 1000, one or more health monitors represented by UEs 1002(1) through 1002(N) operate to monitor various aspects of the health of a user 1004. The health monitors 1002(1) through 1002(N) can provide health information to one or more heterogeneous monitoring systems, represented in system 1000 by application servers 1, 2 (1030, 1032) and UE collection systems 1 through N (1034(1) through 1034(N)). For example, application servers 1, 2 (1030, 1032) and UE collection systems 1 through N (1034(1) through 1034(N)) may be considered third party systems that provide health related recordings of measurements from corresponding health monitors, which may be in the form of a Samsung® Gear Watch or an Apple® iWatch®.
[0085]
[0098] The health information may be provided by various networks. For example, some health monitors may communicate with their monitoring systems over one wireless access network 1022(a), while other health monitors may communicate over another wireless access network 1022(b). In one embodiment, the health monitors may communicate with their respective (e.g., subscribed) monitoring systems over the wireless access network 1022(a) by using a device configured to communicate over the network 1022(a), such as a smartphone, tablet, etc. In one embodiment, the health monitors 1002(1) through 1002(N) provide the health information directly to the health monitoring server 150.
[0086]
[0099] In one aspect, the exemplary system 1000 is Health Insurance Portability and Accountability Act (HIPPA) compliant with physical and technical safeguards included to limit access to the remote health monitoring server to authorized individuals and account holders only. In one example, a unique user ID, automatic logoff procedures, encryption, and decryption are required to access the system 1000.
[0087]
[0100] In one aspect, the remote health monitoring server 150 acts as a measurement result megastore and provides an open platform where multiple data feeds from disparate servers 1030, 1032 and collection systems 1034(1) through 1034(N) are combined. Health data is processed in real time for a large number (e.g., millions) of individuals, such as the user 1004, by a data processing layer that may be part of or in communication with the health monitoring server 150. In addition, the remote health monitoring server 150 may collect data, such as patient records, from hospitals 1040 and care providers 1050. Such data may include medical history, prior treatments, recorded blood tests, and the like. In addition, the remote health monitoring server may collect data from wireless access networks (e.g., 1022(a), 1022(b)) to determine location information of the user 1004 and identify network issues that may prevent data transmission and other variables that affect network connectivity and customer endpoints. In various embodiments, the location of the user 1004 may be determined by various techniques, such as triangulation, multilateration, and / or GPS coordinates, which may be provided by a health monitor (e.g., 1002(1) through 1002(N)) or its corresponding UE (e.g., 1006).
[0088]
[0101] Data from the disparate monitors is analyzed for patterns and trends. Upon determining that one or more thresholds have been exceeded or one or more criteria have been met, the remote health monitoring server 150 can send a notification to the user 1004 via a notification to the user interface of the corresponding health monitor (e.g., UE 1002(1)) or corresponding UE network enabled device 1006. In one embodiment, the health monitor server 150 can send a notification to the relevant care provider 1050 and / or hospital based on the type of condition.
[0089]
[0102] 11 is a block diagram of a data framework of a monitoring system consistent with an example embodiment. The data framework in the system 1100 uses massively parallel processing methods to collect health data in real time from one or more health monitors (UE's). For ease of discussion, the automation layer 1108, intelligence layer 1104, and data processing layer 1102 are shown as separate from the measurement results megastore 1110, but it will be understood that in various embodiments, these functions may be combined on a single platform (i.e., a single measurement results megastore that performs the functions of the data processing layer, intelligence layer, and automation layer).
[0090]
[0103] The data framework includes a data processing layer 1102, an intelligence layer 1104 that may include pattern detection rules 1106, an automation layer 1108, and a measurement results megastore 1110. The types of information received by the data processing layer 352 may include subscriber radio trace (e.g., location) information provided by the UE and / or radio access network, including relevant information from the WEB, medical record information from one or more sources, customer care record information, billing information, application performance monitoring report information, and other information. The data processing layer 1102 may be performed in real-time and / or in batches. Monitoring may be performed continuously, around the clock, at predefined intervals, or in response to a trigger from the monitoring system's intelligence layer 1104.
[0091]
[0104] The intelligence layer 1104 is configured to receive information from the data processing layer 1102 and pattern detection rules 1106 that may be pre-programmed in the intelligence layer. In one embodiment, the detection may be initially programmed by an expert specialist based on prior knowledge. The self-learning function of the intelligence layer operates to detect abnormal conditions without prior knowledge of that particular problem. Specifically, causes and / or KPIs may be detected as present when an undesirable event occurs, such as a dropped call. As described with respect to FIG. 7, a probability tree may be populated and / or updated to store the probability that these causes and / or KPIs are correlated with the undesirable event. In one embodiment, the pattern detection rules 1106 are different depending on the type of patient (e.g., leukemia vs. diabetes). For example, upon identifying a pattern consistent with a health medical condition, the health monitoring system may update its rule set by performing more analysis on health data related to that medical condition to focus on that condition. Thus, the pattern detection rules 1106 may be different for each subscribed user.
[0092]
[0105] To detect a particular condition, multiple data sources processed in the data processing layer 1102 can be analyzed and correlated to determine whether one or more conditions are met. Specifically, the first detected cause and / or KPI are checked against a probability tree (see discussion on FIG. 7) to see if they may possibly be the root cause for the detected undesirable network event, such as a dropped call. The probability tree is then traversed starting from the detected cause and / or KPI. Since the probability tree stores the probability that the detected cause and / or KPI caused the event, the probability tree can be traversed according to the most likely path. In this way, the state of the undesirable network condition can be evaluated for the possibility that the particular cause and / or KPI (and other events / causes along the path in the probability tree) actually caused these undesirable network conditions.
[0093]
[0106] Thus, the intelligence layer 1104 constantly shifts through the data to identify patterns within the disparate data sources. When a predetermined condition is met, the intelligence layer can generate relevant alerts to at least one of the user 604, the care provider 650, contacts previously stored in the Measurement Megastore 1110 database, and / or emergency services. For example, the intelligence layer 1104 can analyze patterns in the data for a particular type of measurement (blood pressure) or correlate different types of measurements or data provided by the data processing layer 352 (e.g., blood glucose, weight, age, white blood cells, etc.) with family history from the Measurement Megastore 1110 to identify any outliers (e.g., meeting a predetermined condition or exceeding a predetermined threshold) that may indicate a health concern with the patient.
[0094]
[0107] Accordingly, the automation layer 1108 can trigger an automatic response to the identified condition of the patient 604 without relying on traditional human processing of the raw data retrieved by the data processing layer. In other words, the intelligence layer coupled with the automation layer 1108 can detect health problems at the subscriber (i.e., patient) level and provide notifications without the subscriber or health care provider even suspecting a health problem with the subscriber. Authorized caregivers, physicians, family and friends, or emergency responders can be automatically contacted based on settings stored in a user profile stored in the measurement megastore 1110. For example, a doctor 211a can provide emergency medical advice. The appropriate entities can be automatically contacted via the automation layer 1108 of the application server of the measurement megastore 1110. The appropriate entities can also be contacted through the wireless access network 122, through the public switched telephone network (PSTN) (not shown), or through an Internet connection (120 in FIG. 1). In one example, different information is sent depending on the receiving entity. For example, if a family member / friend is contacted, location information and the type of emergency may be sent. In contrast, if a doctor is contacted, location information, type of emergency, vital signs, medical history, etc. are sent. Thus, various entities may be contacted to request help, but only specifically authorized entities receive location information. In one example, location information is used to send prescription information to the pharmacy closest to the health monitor (e.g., 102d in FIG. 1).
[0095]
[0108] 12 is a flow diagram of an example process for identifying health conditions using context information received from a measurement results megastore. Consider a subscriber who is wearing one or more health monitors, who believes he or she is healthy. For example, the health monitors can be an electronic fitness bracelet, an electronic necklace that can measure temperature / humidity, and a smartphone that is placed on the subscriber's bed when he or she is resting. Furthermore, the subscriber indicates to an application on the smartphone that he or she would like the application to monitor the subscriber's sleep patterns by placing the phone on his or her bed.
[0096]
[0109] In step 1204, the Measurement Megastore can receive continuous information from all three sensors. For example, the smartphone can provide location information from its global positioning system (GPS) and motion information from its accelerometer, the electronic fitness bracelet can provide heart rate and motion information from its accelerometer, and the electronic necklace can provide measurements of the subscriber's temperature. In various embodiments, the health information may be received by the Measurement Megastore over different wireless access networks and collection systems, as discussed in connection with FIG. 6 above. In particular, the health information may be received by a data processing layer of a data framework that includes the Measurement Megastore.
[0097]
[0110] In step 1206, the intelligence layer of the Measurement Results Megastore analyzes the health information to determine if there is a high level of exercise. For example, the intelligence layer can correlate exercise information from the fitness bracelet with information from the smartphone. If neither source indicates a high level of exercise, the intelligence layer concludes that there is no health problem associated with sleep apnea (i.e., step 1210). However, if one of the sources indicates a high level of exercise that is not confirmed by the second source, the intelligence layer can consult the WEB to determine if the smartphone has a more accurate accelerometer or the fitness bracelet has a more accurate accelerometer and base its analysis of exercise on the more accurate fitness monitor for this measurement.
[0098]
[0111] If the Measurement Megastore intelligence layer identifies a high level of exercise, the intelligence layer then determines whether there is a high heart rate. To that end, in step 1208, the intelligence layer analyzes the health information obtained from the fitness bracelet. In one embodiment, the intelligence layer can consult the WEB (or lookup table) and / or other previously stored data to determine what could be the high heart rate for this subscriber. For example, the stored information can indicate the subscriber's weight, height, and age, as well as previous medical conditions (e.g., pacemaker) that may affect the heart rate. If the heart rate is not considered high for this subscriber, the intelligence layer concludes that there is no health problem associated with sleep apnea (i.e., step 1210).
[0099]
[0112] If the heart rate is determined to be high, in step 1212, the intelligence layer interacts with the automation layer to provide a notification to one or more recipients (e.g., a smartphone, a fitness bracelet, an authorized caregiver, a doctor, family and friends, and / or emergency response) depending on the severity of the condition or the subscriber's predetermined personalization settings. For example, a subscriber may receive a notification on their smartphone indicating that they may be suffering from sleep apnea. The automation layer may also recommend one or more additional health monitors to better analyze the medical condition, and these additional health monitors may be used in subsequent measurements (i.e., step 1214).
[0100] (Example Computer Platform)
[0113] As indicated by the above discussion, the functionality for storing context information, monitoring network elements, remote monitoring of subscriber health, identifying network status based on local knowledge, providing connectivity between network elements, and other functionality may be implemented on computers connected for data communication via the monitoring server 130, the measurement results megastore 140, the eNodeB 119, and network elements of the radio access network 122 operating as components of the radio access network 122, as shown in Figures 1 and 2. While dedicated devices may be used, such devices may also be implemented using one or more hardware platforms that are intended to represent a general class of data processing devices commonly used to execute "server" programming to implement receiving data from disparate sources, retrieving context information, analyzing the information to determine if a predetermined condition is met or a predetermined threshold is exceeded, and providing notifications, as discussed above, despite appropriate network connectivity for data communication.
[0101]
[0114] Figures 13 and 14 provide functional block diagram illustrations of general purpose computer hardware platforms. Figure 13 illustrates a network or host computer platform such as may generally be used to implement a server. Figure 14 illustrates a device having user interface elements such as may be used to implement a personal computer such as computer 102c or a workstation of Figure 1, or to implement a computing device such as a portable handset, a smartphone, a tablet computer, a personal digital assistant, or other type of mobile station or mobile terminal device (e.g., devices 102a through 102d of Figure 1), although the device of Figure 14 may also act as a server when appropriately programmed. It is believed that the general structure and general operation of such equipment as shown in Figures 13 and 14 should be self-explanatory from the high-level illustrations.
[0102]
[0115] For example, a general-purpose computer configured as a server includes a data communication interface for packet data communication. The server computer also includes a central processing unit (CPU) in the form of one or more processors for executing program instructions. A server platform typically includes an internal communication bus, program storage, and data storage for various data files to be processed and / or communicated by the server, although the server often receives programming and data via network communication. The hardware elements, operating systems, and programming languages for such servers are conventional in nature. As used herein, the computer system of FIG. 14 uses massively parallel processing technologies such as Hadoop, Storm, Spark, etc. Of course, the server functions can be implemented in a distributed manner on several similar platforms to distribute the processing load. In various other embodiments, the functions of the monitoring server 130, the measurement results megastore 140, and the remote health monitoring system may be combined in the form of one or more server platforms.
[0103]
[0116] The software functions require programming to retrieve contextual information from the measurement megastore, including executable code and associated stored data, e.g., files used by an application on the computing device to receive data from disparate sources by a processing layer, analyze the data from the disparate sources and its contextual information, and provide notifications. The software code is executable by the computing device. In operation, the code is stored internally to the computing device. At other times, however, the software may be stored in other locations and / or transported for loading into an appropriate computing device system. Execution of such code by the processor of the computing device enables the computing device to perform functions such as retrieving information, analyzing information, providing notifications, etc., in substantial manner as discussed and performed in the implementations shown herein.
[0104]
[0117] Accordingly, aspects of the methods for receiving and processing SUs as outlined above may also be embodied in programming. Program aspects of the technology may also be thought of as "products" or "articles of manufacture" generally in the form of executable code and / or associated data executed or embodied in some type of non-transitory machine-readable medium.
[0105] conclusion
[0118] While the above describes what is believed to be the best mode and / or other examples, it is understood that various modifications may be made therein, that the subject matter disclosed herein may be implemented in a variety of forms and examples, and that the teachings may be applied in a large number of applications, only some of which are described herein. It is intended by the appended claims to claim any and all applications, modifications, and variations that fall within the true scope of the present teachings.
[0106]
[0119] Unless otherwise stated, all measurements, values, ratings, locations, dimensions, sizes, and other specifications set forth in this specification, including the appended claims, are approximate and not exact, and are intended to have a reasonable range consistent with their relevant and customary functions in the art to which they pertain.
[0107]
[0120] Except as immediately set forth above, nothing stated or shown, whether recited or not in the claims, is intended to, or should be construed to, cause a preoccupation to the public of any component, step, feature, object, benefit, advantage, or equivalent.
[0108]
[0121] It will be understood that the terms and expressions used herein have the ordinary meanings as given to such terms and expressions in relation to their corresponding respective fields of inquiry and research, unless a specific meaning is otherwise stated herein. Relative terms such as first, second, etc. may be used merely to distinguish one entity or action from another entity or action, without necessarily requiring or implying any actual such relationship or order between such entities or actions. The terms "comprises", "comprising" or any other variations thereof are intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus that includes a list of elements does not include only those elements, but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element preceded by "a" or "an" does not, without further constraints, exclude the presence of additional identical elements in the process, method, article, or apparatus that includes that element.
[0109]
[0122] The "Summary" of the disclosure is provided to enable the reader to quickly ascertain the nature of the present technical disclosure. It is stated with the understanding that the Summary is not used to interpret or limit the scope or meaning of the claims. Furthermore, in the above "Mode for Carrying Out the Invention", it is noted that various features have been grouped together in various embodiments for the purpose of streamlining the disclosure. This method of disclosure should not be interpreted as reflecting an intention that the claimed embodiments require more features than are expressly recited in each claim. Rather, as the appended claims reflect, inventive subject matter lies in less than all features of a single disclosed embodiment. Accordingly, the appended claims are hereby incorporated into the "Mode for Carrying Out the Invention", with each claim standing on its own as separately claimed subject matter.
Claims
1. A health monitoring server configured to monitor a subscriber's health, comprising: A processor; a network interface coupled to the processor and configured to enable communication over a communications network; a storage device for content and programming; A program stored in the storage device; and execution of the program by the processor includes: receiving disparate data related to the subscriber's health from disparate sources; retrieving the health-related contextual information of the subscriber from a Measurement Megastore; Analyzing the information from the disparate sources and the measurement results megastore to find patterns, trends, or both, the analyzing step including incorporating a root cause identification step by using machine learning techniques, the root cause identification step comprising: clustering the information from the heterogeneous sources and the measurement megastore to identify network events; constructing a probability tree containing the probability that a particular cause is the root cause of a network event; traversing the probability tree and updating the probabilities if a condition is met; and determining a root cause for the network event; the analyzing step comprising: Including, A health monitoring server that constitutes a computer system.
2. A health monitoring server as described in claim 1, wherein a notification is sent to at least one of the heterogeneous sources of the heterogeneous data.
3. The health monitoring server of claim 1 , wherein the disparate data from the disparate sources comprises subscriber wireless traces, medical records from one or more sources, customer care records, billing, application performance monitoring reports, or a combination thereof.
4. The health monitoring server of claim 1 , wherein sources of data include one or more of the subscriber's implanted or worn personal health monitors.
5. 10. The health monitoring server of claim 1, wherein the source of data includes records from a server that stores personal health information of the subscriber.
6. The health monitoring server of claim 1 , wherein the disparate data is initially received from one or more application servers and collection systems across one or more networks before being received by the remote health monitoring server.
7. 10. The health monitoring server of claim 1, wherein the disparate data is received at least in part from one or more health monitors configured to measure heart rate, blood pressure, movement, oxygen saturation, body temperature, blood glucose levels, or combinations thereof.
8. The health monitoring server of claim 1 , wherein the health monitoring server is configured to provide an open platform, where the disparate data from the disparate sources is combined and processed in real time.
9. The health monitoring server of claim 1, wherein a notification is sent to the subscriber, a care provider, a contact person pre-stored in the measurement results megastore database, an emergency service, or a combination thereof.
10. The health monitoring server of claim 1 , wherein execution of the program configures the monitoring server to perform operations continuously.
11. The health monitoring server of claim 1 , wherein execution of the program configures the monitoring server to perform actions at predetermined time intervals.
Citation Information
Patent Citations
Position detection system and its method
JP2007087136A
Medical information system
WO2001075764A1
Communication method and system, access network device, and application server
WO2013152472A1
